The Problem: Agents Transacting Blind
AI agents are increasingly autonomous. They buy data, pay for services, trade on prediction markets, and delegate work to other agents. On Base alone, 251+ x402 services are live, processing millions in USDC micropayments.
But here's the gap: when Agent A is about to pay Agent B $0.50 for a market signal, how does it know Agent B isn't a freshly created Sybil wallet? When a trading bot sees unusual volume on a Polymarket question, how does it distinguish organic interest from coordinated manipulation?
In the human internet, we solved this with credit scores, review platforms, and bank verification. For agents, that layer didn't exist. Until now.
What MoltGuard Does
MoltGuard is an all-in-one trust and integrity API for agents operating on Base. It combines four capabilities that previously required integrating multiple separate services:
1. Agent Trust Score (0-100)
Every wallet on Base gets a trust score based on verifiable on-chain data: wallet age, transaction count, number of unique counterparties, USDC balance, ERC-8004 agent registration, and MolTrust DID cross-reference. Think of it as a credit score for AI agents — computed in seconds from real blockchain data, not self-reported claims.
2. Sybil Detection
Sybil attacks are the agent economy's equivalent of fake Amazon reviews: one operator creates hundreds of wallets to game a system. MoltGuard traces the funding source of any wallet and checks whether the same funder created clusters of sibling wallets. It detects bot-like patterns (many transactions but few counterparties), mass distribution from a single funder, and dormant wallets that suddenly reactivate.
In testing, our scan of a single wallet detected 40 sibling wallets from the same funding source — a cluster that would be invisible to simple balance checks.
3. Prediction Market Integrity
Sportradar monitors FIFA matches for manipulation. MoltGuard does the same for prediction markets — but for AI agents. It checks Polymarket markets for volume anomalies, price-volume divergence, wallet concentration (when a handful of wallets control most of the volume), and suspicious trading patterns. In our scan of 20 active markets, it flagged 10 with integrity concerns.
4. Verifiable Credentials
An agent can request its trust score as a cryptographically signed W3C Verifiable Credential — a portable proof that can be shown to other agents or services. "MoltGuard checked me and gave me score 78" — signed with Ed25519, verifiable by anyone, valid for 7 days. Like a digital certificate that can't be forged.
Architecture
MoltGuard runs on Hono (Node.js) with viem for Base RPC calls. On-chain data comes from Blockscout (free, no API key needed). Agent identity is resolved through the ERC-8004 IdentityRegistry on Base mainnet, cross-referenced with MolTrust's DID registry for off-chain reputation data.
Why Not Just Use Existing Services?
The x402 ecosystem already has individual trust services. Here's what they cover:
| Service | Wallet Score | Sybil | Markets | VCs | MCP |
|---|---|---|---|---|---|
| DJD AgentScore | Yes | — | — | — | — |
| Orac | — | — | — | — | Yes |
| Moltalyzer | — | — | Yes | — | — |
| t54 | Partial | — | — | — | — |
| MoltGuard | Yes | Yes | Yes | Yes | Yes |
Instead of integrating four separate services, an agent calls MoltGuard once.
Endpoints & Pricing
| Endpoint | Price (after EA) | What it does |
|---|---|---|
| /api/agent/score/:addr | $0.005 | Trust Score 0-100 |
| /api/agent/detail/:addr | $0.01 | Full trust report |
| /api/sybil/scan/:addr | $0.01 | Sybil cluster analysis |
| /api/market/check/:id | $0.005 | Market integrity check |
| /api/market/feed | $0.02 | Top-10 anomaly feed |
| /api/credential/issue | $0.05 | Issue trust credential |
All endpoints above are FREE during Early Access. No API key. No signup. Just call them. MoltGuard is x402-ready (middleware installed, prices configured), but the official x402 facilitator doesn't support Base mainnet yet. Once it does, micropayments activate automatically — no breaking change for integrators.
Plus a permanent free tier: /health, /api/info, /api/agent/sample, /api/market/sample, and /api/agent/score-free/:addr (rate-limited).
Try It Now
Technical Details
- Stack: Hono (Node.js), viem, TypeScript
- Chain: Base mainnet (eip155:8453)
- Data sources: Blockscout (on-chain), Polymarket Gamma API, ERC-8004 Registry, MolTrust DID Registry
- Payment: x402 Protocol (USDC on Base) — prices $0.005–$0.05 per request
- Responses: Structured JSON with
_metaobject (data source, timestamp) - Credentials: Ed25519/JWS, W3C VerifiableCredential format, verifiable via
/api/credential/verify - Caching: 5 min for score data, 1 min for market data
- MCP: 7 tools available at api.moltrust.ch/mcp
What's Next
- x402 micropayments go live automatically once the official facilitator adds Base mainnet — no code change for users
- @moltguard/x402-gate npm middleware — let x402 service operators gate access by agent trust score
- OpenClaw Skill for agent-native installation
- x402.org Ecosystem Listing in progress
The goal: every agent in the x402 ecosystem checks its counterparty via MoltGuard before sending payment.
MoltGuard is built by MolTrust (CryptoKRI GmbH, Zurich) — trust infrastructure for the agent economy. DID verification, reputation scoring, Verifiable Credentials, and now integrity monitoring.